post-3333

in case I haven’t said it recently, I’m never doing another corporate lease (or any, for that matter) again. Yes – more screwups with PNC. What kinds? Oh, they added 2000 miles to the final odometer reading and are claiming over $200 in parking tickets that were unpaid that I never got. Lovely. Oh, yeah, and they missed the final payment I made, too.

I hate idiotic corporations that screw things up consistently. The occasional glitch is fine – but things like this speak to a pattern of incompetence. *sigh*

discussions of the day

side threads while I’m hashing out discussions about some Oracle kernel internals were interesting – it focused on religious threads, ranging from analogies between rootkits and the Bible (strange, but a couple of interesting points) to the quantum properties of cesium and (oddly enough) the effects on passages of Genesis.

Very entertaining, but my main thread’s run into a dead end. All I know is that my cache buffer latches are protecting chains that contain 55 buffer heads (pointers to buffers) on average with a low of 35 and high of 107. The painful thing is that I really know where the source problem SQL is and I’d really like to fix it, but it’s largely in need of a redesign at the schema level, and I’m sick of fighting those fights.

post-3330

umm… my _db_block_hash_buckets isn’t prime. In fact, it’s 11*(2^16).

Now, I probably wouldn’t much care, but my cache buffer chains are *long* with lots of contention (x$bh is showing 66 different blocks on some very key segments are all on a single chain, the one with the most contention, and it’s registered over 51k sleeps – ouch). I wanna double it and increment to the next prime (who does hashes with a non-prime number of slots anymore? sheesh), but it’s one of those internal params they swear you have to sacrifice 2 goats and a chicken to change without major threats *sigh*

Ugh – i’ll dig in more tomorrow

giggle

The shared storage failover capabilities include STONITH (“Shoot The Other Node In The Head” — one node physically removes power from the other node for data fencing)